Brep2Shape: Boundary and Shape Representation Alignment via Self-Supervised Transformers

2026-02-07

AI Summary

This work addresses the representational gap in CAD between boundary representations (B-reps) and intuitive shape descriptors—where continuous methods offer geometric precision but lack intuitiveness, while discrete approaches are intuitive yet suffer from limited accuracy. To bridge this divide, the authors propose a self-supervised pretraining method that aligns these two representation paradigms by predicting dense spatial points generated from parametric Bézier control points. A dual-stream Transformer architecture is introduced, separately encoding surfaces and curves, and augmented with a topological attention mechanism to preserve both geometric fidelity and topological consistency. This approach achieves, for the first time, self-supervised alignment between B-reps and shape-based representations, significantly improving convergence speed, accuracy, and scalability across multiple downstream tasks, and establishing state-of-the-art performance.

Technology Category

Computer Vision: Representation Learning for VisionKnowledge Representation and Reasoning: Geometric, Spatial, and Temporal ReasoningMachine Learning: Representation Learning

Application Category

User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendationGraph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphsSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Boundary representation (B-rep) is the industry standard for computer-aided design (CAD). While deep learning shows promise in processing B-rep models, existing methods suffer from a representation gap: continuous approaches offer analytical precision but are visually abstract, whereas discrete methods provide intuitive clarity at the expense of geometric precision. To bridge this gap, we introduce Brep2Shape, a novel self-supervised pre-training method designed to align abstract boundary representations with intuitive shape representations. Our method employs a geometry-aware task where the model learns to predict dense spatial points from parametric B\'ezier control points, enabling the network to better understand physical manifolds derived from abstract coefficients. To enhance this alignment, we propose a Dual Transformer backbone with parallel streams that independently encode surface and curve tokens to capture their distinct geometric properties. Moreover, the topology attention is integrated to model the interdependencies between surfaces and curves, thereby maintaining topological consistency. Experimental results demonstrate that Brep2Shape offers significant scalability, achieving state-of-the-art accuracy and faster convergence across various downstream tasks.

Problem

• Boundary representation
• Shape representation
• Representation gap
• CAD models
• Geometric precision

Innovation

• B-rep
• self-supervised learning
• Dual Transformer
• topology attention
• geometric representation